2014
2014-05-20
How Voters Make up Their Minds:
Consideration Set Models for Party
Choice in European and National
Elections.
Maria Oskarson, Henrik Oscarsson
& Edvin Boije
Report 2014: 03
Swedish National Election Studies Program
Department of Political Science
University of Gothenburg
***** FIRST DRAFT - DO NOT QUOTE ****
How Voters Make up Their Minds: Consideration Set
Models for Party Choice in European and National
Elections.1
Paper prepared for the Swedish Network for European Studies in Political Science
Conference: ELECTIONS AND DEMOCRACY IN EUROPE, Brussels, 7–9 April 2014
Maria Oskarson, Henrik Ekengren Oscarsson, and Edvin Boije
Swedish National Election Studies Program (SNES), Department of Political Science
University of Gothenburg, Sweden
Contact: [email protected]
Abstract
In order to gain more insight into voters’ decision-making processes, we applied a consideration set approach to explain party choice in European and national elections. In line with
the work on first and second decision rule criteria in EP elections by Catherine de Vries and
her co-authors (C. d. Vries, Steenbergen, & Hangartner, 2009), we set out to develop some
first hypotheses about how standard components of voters’ calculus have different weights
in the two types of elections and at different stages of the voting decision process. We
present preliminary tests of hypotheses about consideration set size and content, primary
and secondary decision criteria, and the effects of proximity voting using Swedish data from
two self-recruited internet campaign panels conducted in conjunction with the 2009 EP
election and the national election in 2010. The findings indicate that, as expected, the
impact of left–right ideology is lower in European elections than in national elections, both
as a primary and secondary decision rule. At this stage, we have not been able to discern
whether this is an effect of cross-pressure from the ideological position to the EU, which is
probable. We conclude that the consideration set model approach is a viable way of
disentangling the decision processes behind party choice in European parliamentary
elections, as well as national elections.
1
This paper presents preliminary hypotheses and “pre-tests” for the research project “Developing
Consideration Set Models of Voting Behavior” run by Henrik Ekengren Oscarsson and Maria
Oskarson, with Edvin Boije as research assistant. The project is financed by Riksbankens
jubileumsfond for the years 2014–2016. Project number P13-0721:1.
1
Introduction
The most fundamental conception of elections to the European Parliament is that they are
“second order elections” (Marsh, 1998; Reif & Schmitt, 1980; Schmitt, 2005). Compared to
national elections, which are “first order,” elections to the European parliament are
perceived as being less important as less is at stake, especially since no government will be
formed on the basis of the electoral outcome. Due to their second order status,
participation is generally lower; big parties as well government parties tend to lose support,
whilst smaller, niche, and fringe parties tend to succeed in comparison to their performance
at national elections. Also, multiple studies have concluded that the voters in European
elections base their voting decision on an evaluation of national government performance
rather than on European issues. However, since this conception was formed, the European
Union, as well as the power and significance of the European Parliament, have changed.
Even though the general characterization of the European Elections as second order is still
valid, recent research has pointed to the fact that with more information on European
matters, voters now to a higher degree base their voting decision on actual EU matters (Sara
B. Hobolt & Spoon, 2012; Sara Binzer Hobolt & Wittrock, 2011; C. E. d. Vries, Brug, Egmond,
& Eijk, 2011).
To cast a vote for a party in an election—local, national or European—is a manifestation of
two decisions: first, whether or not to vote; second, in the event of voting, which party to
vote for. How voters make these decisions has of course been the focus of research on
voting behavior and elections ever since the beginning of individual-based election research.
Still, the knowledge of the actual decision-making process behind party choice is limited.
That factors such as social background, political identification, degree of political
sophistication, ideological position, and perceived issue proximity are central to party choice
is well documented (R. J. Dalton & Klingemann, 2007). It is also well documented that
electoral volatility has increased over time, as long-standing stabilizing factors such as
political identification and class voting have waned (Bengtsson, Hansen, Narud, Hardarson,
& Oscarsson, 2013; R. Dalton, 2013). Thus, we can expect that there is, behind the party
vote, increasingly an actual party choice rather than a manifestation of any long-lasting
party identifications. However, research on how voters actually make up their minds has still
not been able to fully detect how different explanatory factors behind the decision how to
vote relate to each other and when they occur in the decision-making process.
In order to respond to this development we applied the “consideration set models of party
choice” (CSM) as an intriguing remedy to the many challenges of a highly individualized
voting behavior. The CSM approach will be illustrated empirically and tested by elections in
Sweden—national as well as European. The focus of this paper is on how a CSM approach
2
can further clarify the decision-making process in secondary elections (i.e., European
elections) compared to primary elections (national elections).
Theoretical approach
The general idea behind consideration set models of party choice is that, at the time of
elections, many voters actively consider voting for more than one party. But, voters are not
believed to consider the entire supply of parties, at least not in multiparty systems. Rather,
voters are believed to enter election campaigns with an existing subset of voting
alternatives from which they subsequently select the party to vote for. While earlier
generations of citizens often manifested group-based interests or identifications by
routinely supporting certain parties at elections, many voters of the 21st century are
believed to engage in an actual decision-making process. This decision process takes place in
a context of intense campaigning and information processing. In spite of this development,
party choice still tends to be analyzed with the same models and methods used in more
predictable and steady periods. The CSM approach, which is inspired by consumer research,
provides a theoretically-based explanatory model in response to the contemporary voting
behavior of citizens.
Applying a CSM approach to party choice enables researchers to combine insights from
political sociology with insights into political psychology. More precisely, CSM allows a
thorough analysis, and takes into account how voters expose, evade, and process political
information. The implication of the CSM assumption that voters form consideration sets to
reduce the cognitive costs of resource-demanding information processing is that voters only
expose themselves to certain political information during intense elections campaigns,
namely, the type of information that is needed to pick a winner from a preselected set of
considered alternatives. These assumptions of the model have testable implications, as we
would expect that exposure to party messages and campaign events will be limited to the
parties included in an individual’s party set.
When phased decision-making processes are studied in consumer research, more stages are
often discussed. For example, according to Shocker, Ben-Akiva, Bocara, & Nedungadi (1991)
as well as Roberts and Lattin (1997) the decision process involves several nested steps. All
possible options form the universal set, whilst the subset of options that the individual
actually knows about forms the awareness set, of which some options are considered and
others not. Thus, the consideration set is nested in the awareness set (Roberts & Lattin,
1991, 1997; Shocker, Ben-Akiva, Bocara, & Nedungadi, 1991). In previous studies of the
application of a consideration set approach to party choice the first two stages, the
universal set and the awareness set, are generally excluded since the universal set (all
possible parties) still presents a more restricted number of options than is the case in most
consumption research. For this reason there is also believed to be only very limited variation
3
between voters in the awareness set in established party-based democracies, which is why
the analysis of this stage is considered futile and the use of the consideration set model for
party choice generally only distinguishes two separate stages in the decision process: the
consideration stage and the choice stage (Steenbergen & Hangartner, 2008; Steenbergen,
Hangartner, & Vries, 2011; Wilson, 2008; C. d. Vries et al., 2009). The consideration set
model approach for party choice is schematically illustrated in Figure 1.
Awareness set
Consideration set
Party choice
P1
P2
P2
P4
Choice stage
P4
Consideration stage
P3
P1
P4
P5
P6
TIME
Election campaign starts
Election day
Figure 1. Schematic illustration of the decision-making process for party choice according
to the consideration set model. Note: P1–P6 represents the parties entering the election.
The first stage, when the focus is on party choice, is accordingly the formation of the
consideration set, in other words, concerning which parties there is considered to be an
option to vote for and which parties are excluded early in the election campaign or have
already been excluded even before the campaign actually starts. Of course, some voters
have only one party in their set and might see it as excluded if they voted for any other.
However, according to research on the decline of party identification and voting stability,
this group is believed to be in decline. Still, the size (number of parties) and content (which
parties) of the consideration set is expected to be mainly dependent on more stable and
longitudinal predispositions such as ideological orientation and proximity, political
4
identification, political sophistication, and voting history. In the second stage an actual
decision about which party to vote for is formed. This decision process is limited to the
parties in the consideration set, and is dependent mainly on short-term factors such as issue
positions and issue proximity, evaluations, and expectations of party performance, and
events taking place during the election campaign (Steenbergen & Hangartner, 2008; Wilson,
2008; C. d. Vries et al., 2009).
Distinguishing the two stages—the consideration stage and choice stage—from each other
requires more finely-tuned analytical models, where it is possible to distinguish factors and
criteria that influence the consideration set formation from factors and criteria influencing
the actual vote choice, and it also requires the temporal factor of different criteria and
considerations to be examined. Furthermore, this is consistent with the view of voters as
“cognitive misers” who use heuristics to minimize information costs (Popkin, 1991;
Sniderman, Brody, & Tetlock, 1991; Wilson, 2008). To distinguish the consideration set from
the actual choice set is also in line with Zaller’s notion that voters’ attentiveness and
response to election campaigns are expected to be dependent on their predispositions, not
only in ideological terms, but also in terms of what parties they actually consider voting for
(Zaller, 1992). The distinction between separate phases in the decision-making process
about party choice enables us to distinguish the effects of election campaigns in a more
realistic fashion than when we simply understand party choice as a choice between all
parties in the party system.
So far, research on applying consideration set models to party choice has been limited and is
mainly oriented towards statistical modeling, for example, probabilistic choice set
multinomial logit models (Paap et al., 2005; Steenbergen & Hangartner, 2008) or choice set
logistic regression (Steenbergen et al., 2011). In part, the technical nature of previous
research into consideration set models of party choice has been due to the limited
instrumentation in surveys, necessitating various indirect techniques to identify the
considerations sets. A more complete analysis is presented by Carole J. Wilson, who shows
that local party strength forms a contextual determinant of the consideration of voters
(Wilson, 2008).
Previous research with consideration set models in Sweden
Previous research that has applied to consideration set approaches to voter choice in
Sweden has been concentrated on issues of survey instrumentation for identifying choice
sets. There are a few descriptive analyses of choice set size (i.e., the number of parties
voters consider voting for) and choice set content (i.e., what combination of parties are
selected in the party formation stage) (Oscarsson, 2004, 2009; Oscarsson, Gilljam, &
Granberg, 1997; Oscarsson & Holmberg, 2008). A large number of ways of identifying the
choice sets of individuals have been tested out using both established measures such as
5
dislike–like scales, feeling thermometers, propensity to vote scales, and second best party
preferences, and new survey instruments with questions about the likely, feasible, and
considered alternatives and what parties’ voters can choose between. The choice of the
operationalization used in this study to identify considered alternatives rests heavily on past
experiences of applying different instruments in this context.
However, CSM has also been applied in explanatory analyses of party choice. The approach
allows for voter, party, and choice set characteristics to be modeled simultaneously. A pilot
study by Oscarsson (2004) retrieved consideration sets using the number of parties that
were positively evaluated along an eleven-point dislike–like scale from −5 to +5, using data
from the Swedish national elections 1979–2002; it estimated the effects of party size,
campaign momentum, leader evaluations, ideological proximity, perceived party
competence, and habit (previous voting for the party) on the probability of a given party
being picked in the final stages of party choice. In these analyses, leader evaluations—the
party picked from a set was the party with the most liked leader—had a surprisingly small
and insignificant effect while momentum, habit, and party size turned out to be significant.
Ideological proximity and general assessments of parties’ policies on prioritized issues had
the largest and most robust impact on the probability of voting for a party included in a
voter’s choice set.
In another study, open-ended questions about the pros and cons of all the alternatives in
the choice set was used to assess the selection criteria applied by voters in the final stages
of the decision process (Oscarsson & Holmberg, 2008). The results of this explorative
approach showed that among the open-ended answers, statements about ideology and
issues were most frequent. In second place came voters’ judgments of confidence and trust
in the alternatives. References to group loyalty and socialization were only found among the
left-green party voters, and references to the strategic context of the election (parties’
standing in opinion polls, government formation, and parliamentary thresholds) were
present to a surprisingly large degree.
Swedish election studies have introduced appropriate questions into recent surveys and
internet campaign panels carried out in conjunction with both national and European
elections. This has provided a unique opportunity to analyze consideration set models of
party choice more thoroughly. More specifically, we are now able to empirically test a large
number of hypotheses about consideration set formation in low and high stimuli elections.
Although other data sources are readily available, in this paper we focus on analyzing data
collected through opt-in internet campaign panels in 2009 and 2010 (Dahlberg, Lindholm,
Lundmark, Oscarsson, & Åsbrink, 2011). The justification for using campaign panel data is
that the comparison of the decision processes taking place in a European parliamentary
election and a national election demands that data are available that is collected before the
6
elections. Pre-election data on consideration set formation is not available in the large
representative Swedish EP election studies.
Party choice in European Parliament elections compared to national elections
When it comes to comparing voting decisions in European elections to national elections,
the consideration set approach has the potential to offer a deeper understanding of the
mechanisms at play when it comes to party choice in second-order compared to first-order
elections. Cleavage structures and ideological orientations are based historically in national
politics and believed to structure the relation between voters and parties; accordingly, they
provide the main explanation of a voter’s consideration set in elections. These fundamental
structures in the relationship between voter and party do not have a counterpart in the
European elections, restricting the possible impact of European issues on the discrimination
between the parties already in the consideration set.
In a paper from 2009 on EU issue voting in Britain from a consideration set perspective,
Vries et al. (2009) built on the work on issue evolution by Carmines and Stimson (1986) and
asked if EU issues act as primary (consideration stage) or secondary (choice stage) decision
criteria when voters form their consideration sets. The basic assumption was that in the
case of gradual issue evolution an issue starts as a “secondary decision criterion,” or put
differently, the issue is not of such great importance that it is used to eliminate alternatives.
Over time, such an issue may become a “primary decision criterion,” decisive in determining
which parties people consider voting for (Carmines & Stimson, 1989; C. d. Vries et al., 2009,
p. 8). The results suggest that even though EU issue voting has increased in Britain, it still
does not have the position of primary decision criteria. Furthermore, there are grounds for
believing that the issue evolution process could depend on the level of information on EU
matters, since the degree to which EU issues actually influence decisions on party choice in
the European Parliament elections is found to depend upon information about EU issues.
Vries et al. (2009) found in a comparative study of the 2009 election that EU issue voting
was slightly more pronounced among sophisticated voters, but it was clearly more
determining in contexts that provided more information on EU matters (C. E. d. Vries et al.,
2011). Accordingly, it is probable that higher levels of sophistication and information
increase the occurrence of EU issues as secondary decision criteria.
By applying a CSM perspective on party choice for the European Parliament elections, we
will be able to better understand and depict the underlying decision-making processes
among voters. In order to do this, we will make distinctions between the consideration
stage and the choice stage. For the consideration stage, it is also vital to distinguish between
the size of the consideration set (how many parties) and the content of the set (which
7
parties). Together with the insights from previous research on European Parliament
elections as second order elections, a number of hypotheses will then be formulated.
First, in considering the lower intensity and visibility of the EUP elections, we expected that
more voters would actually report consideration sets of more than one party, and also that
the initial consideration sets are in general larger (contain more parties) in EP elections than
in national elections. Consequently, two descriptive hypotheses can be formulated:
H1)
There are more voters with consideration sets of more than one party in EP
elections than in national elections
H2)
Voters with consideration sets have larger sets (more parties) in EP elections than in
national elections
Maybe, the most debated and researched aspect of voting in the elections to the European
parliament is the question of whether the party choice is based on European issues, or on
the general left–right orientation and national issues (van der Brug & Eijk, 2007). Due to the
cleavage character of the Swedish multiparty system, and the low salience of the European
elections we believed EU issues would have an impact as secondary decision rules (choice
stage), whilst the probability of even considering voting for a party is primarily dependent
on the left–right ideological position. Furthermore, even though national elections matter
for European integration through the government’s influence in the Council, European
affairs are even less politicized and discussed in national election campaigns. In spite of the
low perceived significance of EU issues in European elections, it is still viable to expect EU
issues to matter more at the choice stage of the EP elections than in the national elections.
Accordingly, we have formulated two hypotheses on the relative impact of the left–right
and European ideological positions:
H3)
The content of the consideration set for the European Parliamentary election is
dependent on left–right position rather than the position on EU issues.
H4)
The position on European issues matters more at the choice stage (which party to
vote for) than at the consideration stage.
Still, to a large degree the weak impact of the position on the EU dimension has been found
to be due more to the alternatives presented by the parties on EU issues than a lack of EU
attitudes among the voters (van der Eijk & Franklin, 2007; C. E. d. Vries, 2007). Therefore, it
is not just a question of how the voters place themselves ideologically that matters, but also
how the party positions are perceived, that is, the ideological proximity between voters and
parties. Referring to the previous discussion of primary versus secondary decision criteria,
we expected the consideration stage to reflect that:
8
H5)
The consideration set is based more on left–right proximity than on EU issue
proximity in EUP as well as national elections.
H6)
The effect of ideological left–right proximity on inclusion in the consideration set is
lower in EP elections than in national elections.
Furthermore, in the choice stage we believed that:
H7)
Proximity voting along the left–right dimension is lower in EP elections than in
national elections.
H8)
Proximity voting on the position on European integration is higher in EP elections
than in national elections.
However, due to the lower intensity of EP campaigns, political sophistication is expected to
have stronger interaction effects in the EP than in the national elections, since it is believed
to be more cognitively demanding in a low intensity campaign (C. E. d. Vries et al., 2011).
H9)
Political sophistication has a stronger interactive effect on proximity voting in EUP
elections than in national elections.
One aspect of the EP elections as second order elections is that less is at stake. This leaves
the floor open for voting for a “best” party, disregarding issues on government formation or
tactics. Accordingly, we can formulate a final hypothesis regarding the choice stage as
follows:
H10)
Voters who consider voting for more than one party tend to vote more for the most
sympathized with (dislike–like scale) party in the EP elections than in the national
elections.
Data
The Swedish election study program offers good opportunities to explore these hypotheses.
Empirical data on Swedish voters choice sets has been accumulated over the past ten years
and a number of different suggestions for instrumentation have been proposed (Holmberg
& Oscarsson, 2004; Oscarsson, 2009; Oscarsson et al., 1997; Oscarsson & Holmberg, 2008).
However, ordinary election studies are not the most appropriate for enabling analyses of
consideration set models since they are restricted to measurements at one or, at the most,
two points in time, and they also not always include optimal instrumentation for
consideration sets for party choice. In order to fully study the decision process we need
panel studies with more steps. The National Swedish Election Studies (SNES) program has
conducted web-based campaign panel studies in conjunction with the 2002 and 2006
national elections, the 2004 referendum on the euro, and the election to the European
9
Parliament in 2009. These studies were quite limited in scope, but have opened up
opportunities to develop designs with web-based campaign studies. During the campaign
for the 2010 national election, a more elaborate campaign panel was designed as part of the
research program entitled, “Multidisciplinary Research on Opinion and Democracy” (MOD)
at the University of Gothenburg. For the 2014 elections to the European Parliament, as well
as to the national parliament, a specific web-based campaign panel was set up explicitly for
the research project, “Developing Consideration Set Models of Voting Behavior.” 2 The
analyses in this paper are meant to form a pre-test for analyses of the consideration set
model for the 2014 elections, and will be based on the 2009 and 2010 internet campaign
panels. Most of the previous analyses have been largely descriptive. In the coming years, we
will have the opportunity to exploit the full analytic power of existing data, and cover the EP
and national elections in 2014 for the purpose of developing the CSM.
The 2010 internet campaign panel as well as the 2009 European election internet campaign
panel both include explicit questions about party choice considerations, enabling the
consideration set models proposed here to be analyzed. The 2009 campaign panel study
was a continuation of the 2006 study, using the same self-recruited sample. It was a fivestep panel, with four questionnaires applied before the election and one after. The 2010
internet campaign panel was a six-wave panel of initially 14,434 self-recruited respondents
(Dahlberg, Lindholm, Lundmark, Oscarsson, & Åsbrink, 2010). The survey was carried out
between 24 August and 30 September, which was before and just after the national general
election, which was held on 19 September. The recruitment was performed through ads and
pop-up ads on homepages linked to the Department of Political Science, University of
Gothenburg, as well as through cooperation with several newspaper web editions. Although
these campaign panels have been based on convenience samples (self-recruited via social
media channels), it is possible to track individual level changes during the intense election
campaigns in 2009 and 2010. Due to the self-recruited character of the panel it is not
representative of the Swedish population as a whole—men, middle-aged, highly educated,
and politically interested respondents are overrepresented.
Operationalization and instrumentation
Most previous studies applying consideration set models to party choice have been
restricted to basing their analyses on questions not specifically designed for this purpose,
such as dislike–like scales (Oscarsson, 2004). To infer consideration sets from dislike–like
scales requires the researcher to define some kind of more or less arbitrary cutoff point in
2
The projected is run by Henrik Ekengren Oscarsson and Maria Oskarson, with Edvin Boije as
research assistant. The project is financed by Riksbankens jubileumsfond in the years 2014–2016.
Project number P13-0721:1.
10
order to identify a consideration set, for example, all parties with a positive value. However
this cutoff is done, we can never be sure that the respondent actually has the same
considerations in voting as for parties that are “liked.” When it comes to voting, factors such
as coalition possibilities, strategic voting, and short-term issue salience might also matter in
which parties are considered. Another strategy has been to use recall questions in postelection surveys (Did you consider voting for any other party?). Even though this strategy
might lead to the actual choice set on Election Day, it is not possible to use it for tracking the
actual decision process that lies behind the party choice.
In the Swedish web panel surveys used here, the questions are explicitly designed for
analyses of pre-election consideration sets. The question used is: Which party or parties are
you considering voting for in the upcoming (national/European) elections? Is there any
other party you are considering voting for? The questions were asked in the first phases of
the panel surveys, a month or so before the election in question, and then in consecutive
surveys. In order to analyze how different factors such as ideological proximity affects
whether or not a party is included in the consideration set we “stacked” the dataset. This
means that every alternative within the possible consideration set, here the party system, is
used as the unit of analysis; in other words, every respondent occurs as many times as there
are parties to choose between. For the analysis with the stacked datasets, robust logistic
analysis was employed. The standard errors were corrected since the observations
emanated from the same respondent and are not independent.
Results
The first two hypotheses concern the consideration sets per se, and how they varied
between the 2009 European election and the National election in 2010. The expectation
was that the second-order character of the EP election would be reflected in the weaker
influence of long-term factors such as party identification, and accordingly voters would
have larger consideration sets in the European elections, as formulated below. Due to the
non-representative traits of the web panels on which the analysis was based, our focus of
interest was on the differences between the European and the national elections, rather
than on the absolute levels. Hence,
H1)
There are more voters initially with consideration sets of more than one party in EP
elections than in national elections.
H2)
Voters with consideration sets have larger sets (more parties) in EP elections than in
national elections.
11
To test the hypotheses, we specified the “consideration stage” as 21 days before the
respective elections. Table 1 shows that the proportion of respondents in the panels
actually considering more than one party was very similar in the 2009 European Parliament
election (55%) and the national parliamentary election of 2010 (58%).
Table 1. Share of respondents with more than one party in a consideration set, 21 days
before election
Swedish EP election
Swedish national election
(EP 2009)
(EP 2010)
Considering one party
45%
42%
Considering >1 party
55%
58%
Number of
respondents
1650
1918
Note: Respondents not reporting any party are not included in the percentage base.
In Table 2 we report the number of parties in the consideration sets at the respective
elections from those respondents who were considering more than one party. The
difference in the average size of the consideration sets is negligible: 2.62 parties in 2009
compared and 2.65 parties in 2010.
Table 2. Average consideration set size among respondents considering more than one
party, 21 days before election
Swedish EP election
Swedish national election
(EUP 2009)
Average consideration set size
2.62
(EP 2010)
2.65
Number of respondents
901
1119
Note: Respondents not reporting any party are not included in the percentage base.
So, these preliminary results suggest that there are no significant differences in either the
proportion of the voters that were considering more than one party three weeks before the
election or in how many parties the voters were considering. In the Appendix (Tables A1 and
A2), we report the sizes and content of the consideration sets of the internet panel samples
in more detail. The similarity between the consideration set for the European (EP) and the
national (Riksdag ) election is further validated in this more detailed analysis. The main
difference is that in the national election the left bloc included the Social Democratic Party,
the Left Party and the Green Party as the most common three-party consideration set,
whilst the one-, two-, and three-party consideration sets for the EP election were
12
dominated by the right-wing coalition parties. Of course the content of the consideration
sets reflects the biased self-recruited sample and is not representative of the Swedish
electorate. In the following, we have therefore concentrated on ideology as an explanatory
factor for the consideration and choice stages, and the comparisons between the two
elections.
The first hypotheses concerns the impact of the respondents’ ideological position on the
content of the consideration sets in the European election of 2009.
H3)
The content of the consideration set for the European Parliamentary election is
dependent on left–right position rather than the position on EU issues.
H4)
The position on European issues matters more at the choice stage (which party to
vote for) than at the consideration stage.
In order to test these hypotheses we looked at the effect of ideological position on the
inclusion of the party in the consideration set, compared to the effect of ideological position
on actual voting for the party. This means that the analysis was made party by party, as
reported in Table 3 below.
13
Table 3. The effect of left–right ideological position compared to position on EU
integration on consideration set and on party vote in the European Parliamentary election
of 2009; logistic regression
Consideration
stage
Choice stage
L
S
G
C
Lib
CD
M
LR (0–10)
−0.86***
−0.60***
−0.39***
0.23***
0.27***
0.39***
0.83***
EU (0–10)
−0.17**
0.23***
0.14***
0.27***
0.30***
0.08
0.27***
Constant
2.88***
0.68**
0.23
−5.01***
−4.53***
−5.52***
−7.78***
Log-likelihood
−264.12
−427.91
−517.86
−390.77
−575.42
−324.59
−487.76
Pseudo R2
0.39
0.24
0.12
0.07
0.11
0.09
0.36
N
1161
1161
1161
1161
1161
1161
1161
LR (0–10)
−0.96***
−0.53***
−0.35***
0.21**
0.28***
0.33***
0.66***
EU (0–10)
−0.41***
0.12
0.17**
0.08
0.19***
0.19*
0.26***
Constant
3.04***
0.26
−0.93**
−5.28***
−4.57***
−6.36
−7.88
Log-likelihood
−132.71
−298.40
−321.49
−141.28
−403.91
−179.22
−361.27
Pseudo R2
0.41
0.19
0.09
0.03
0.07
0.07
0.23
N
1011
1011
1011
1011
1011
1011
1011
Comment: The party abbreviations are L = Left Party, S = Social Democrats, G = Green Party, C = Center Party, Lib
= Liberal Party (People’s Party), CD = Christian Democrats, and M = Moderate Party (Conservative). Dependent
variable for the consideration stage is “the party is included in the consideration set 21 days before the EP
election.” Dependent variable for the choice stage is the party vote in the EP election of 2009. The independent
variable LR is the respondents reported position on a left-right scale where 0 = extreme left and 10 = extreme
right. Ideological position on European integration is based on an index constructed from respondents’
3
evaluation of EU impact on a number of policy areas. The index ranges from 0 = most EU negative to 10 = most
EU positive (Cronbach’s alpha: 0.87). Marked cells are not in accordance with the hypotheses.
The table indicates no unequivocal support for either hypothesis, but a rather different
patterns for different parties. The first hypothesis concerned whether the parties for which
voting is being considered (part of the consideration set) are mainly influenced by the
ideological left–right position of the respondent. This is not the case for the Center Party or
the Liberal Party, for which the impact of EU position is higher than the impact of LR
position. However, for most of the other parties the effect of EU position is significant, if not
stronger than LR position. Only for the Christian Democrat Party (CD) do we have an
insignificant effect of position on the EU dimension.
3
The question was formulated as: “What is your opinion on the impact of the EU membership on the
following areas?” The policy areas are environment, economy, employment, health care,
immigration, social welfare, and the possible impact on EU politics. The alternatives ranged between
1 = big improvement and 7 = large deterioration.
14
The second hypothesis on the impact of ideology on the EP election of 2009 was that the EU
position should matter more for the party vote than for inclusion at the consideration stage.
Here, the result varies somewhat between the parties. For the Centre Party and the Liberal
Party, the impact of EU position is far weaker on party choice than on inclusion in the
consideration set, and for the Centre Party the effect for party choice is insignificant. Also,
for the Social Democrats the effect of EU position is insignificant for party choice, whilst it
has a significant effect in the consideration stage, albeit weaker than that for left–right
position.
As discussed previously, ideological proximity is generally believed to be one of the main
factors influencing which parties people consider voting for. Not least, in a multiparty
system such as the Swedish, several parties might be perceived as quite close ideologically.
We therefore proceeded to compare the effect of ideological proximity on the content of
the consideration sets in the two elections. Unfortunately, there were no questions about
perceptions of the EU ideological position of parties in the surveys, which means that any
comparison between EU proximity and left–right proximity as primary decision rules must
be deferred at this stage.
However, it is possible to test the effect of left–right ideological proximity on the
consideration stage using the 2009/2010 data. The following hypothesis is based on the
assumption that even though previous research has stated that voting in European
parliamentary elections is primarily influenced by national left–right issues, the left–right
position should matter less in the EP elections due to the fact these elections are also
influenced by the EU dimension. Accordingly H6 stated that:
H6)
The effect of ideological left–right proximity on inclusion in the consideration set is
lower in EP elections than in national elections.
The hypothesis was tested with robust logistic regression with the stacked datasets from
2009 and 2010 combined. The dependent variable was whether a party is part of the
consideration set. The independent variable is ideological proximity between the
respondent and the party in question, operationalized as the difference (in absolute
numbers) between the respondents’ positions on a left–right scale and the perceived
position of the party. The results are displayed in Table 4 and Figure 2 below.
15
Table 4. Probability of the party being included in the consideration set as a function of
ideological left–right proximity; robust logistic regression
Ideological proximity
National election
National election*
proximity
constant
N
Log pseudolikelihood
Pseudo R2
Coefficients
−0.53***
0.46***
−0.09**
Robust standard error
0.020
0.085
0.036
−0.21***
0.043
2156
−6138.6417
0.18
Comment: Combined stacked datasets (voter–party dyads) from the internet campaign panels in
2009 and 2010. The dependent variable is whether the party is included in the consideration set or
not. The 2009 European Parliament election is the reference category, and the 2010 national election
is a dummy variable. Proximity is the absolute difference between a respondent’s position and the
perceived party position on the left–right scale (0–10). Significance levels are marked *= <0.05, ** =
<0.01 and ***=<0.001.
SNES Internet Campaign Panel 2009 & 2010
0
Pr(party in consideration set)
.4
.2
.6
Predictive Margins of EUP with 95% CIs
0
1
2
3
4
5
6
LR proximity
RD 2010
7
8
9
10
EUP 2009
Figure 2. Probability of the party being included in the consideration set as a function of
ideological left–right proximity.
The results support the hypothesis that ideological left–right proximity has a stronger effect
on the consideration of parties in the national election than in the election to the European
16
Parliament, as the difference between the two curves is significant. The figure also supports
the notion that ideological proximity is a vital factor in party considerations in both
elections. The probability of having a party in the consideration set basically doubles if there
is a tie up between the respondent’s owns position and the perceived position of the party,
compared to a situation where there is a two steps difference in ideological placement.
Unfortunately, we are not able to contrast these results with the effect of ideological
proximity along the ideological dimension by focusing on European integration, as the
questions were not asked in the 2009 or 2010 panel. The hypothesis that proximity voting
on the position on European integration is higher in EP elections than in national elections
will therefore have to wait for the upcoming analyses of the 2014 panel study. This is also
the case for the hypothesis concerning the mediating effect of political sophistication.
After concluding that left–right ideology is of higher significance in considering a party for
the national than the European elections, the next step was to test if this is also the case at
the choice stage; that is, when the voter makes the final voting decision from the different
parties being considered. The hypothesis (7) is that due to the second order character of the
European elections as well as the impact of attitudes to European integration, left–right
proximity should matter less than in the national elections. Table 5 and Figure 3 provide the
results.
Table 5. Left–right (LR) proximity voting among voters with the consideration sets in the
2009 EP election and the 2010 national election; robust logistic regression
LR proximity
National election
LR proximity* National
election
Constant
Number of observations
Log pseudolikelihood
Pseudo R2
Coefficients
−0.53*
0.37**
−0.16*
Robust standard error
0.33
0.10
0.06
−0.85***
0.05
737
−1627.4495
0.15
Comment: Combined stacked datatsets from E-panels for 2009 and 2010, meaning that every
respondent occurs as many times as there are parties to choose between. Only respondents with
more than one party in the consideration set are included. The 2009 European election is the
reference category, and the 2010 national election is a dummy variable. Proximity is the absolute
difference between respondent’s position and the perceived party position on the left–right scale (0–
10). Significance levels are marked *= <0.05, ** = <0.01 and ***=<0.001.
17
Figure 3. Left–right proximity voting among voters with the consideration sets in the 2009
EP election and the 2010 national election.
The results are in line with the hypothesis that left–right proximity should matter more for
the final party choice in the national elections than in the European. The effects are slightly
weaker at the choice stage (here) than in the consideration stage presented in Table 3 and
Figure 2.
As discussed earlier, a central difference between the national and the European election is
that in the European election there is no issue of government formation, and consequently
less focus on potential coalitions and a credible government. This has led to the expectation
that party choice in the European Parliament election is based more on “true” party
sympathy, which is free from strategic or instrumental considerations. With the
consideration set approach this would leave us to believe that among people considering
choosing between several parties, the actual sympathy for the parties matters more in the
EP elections than in the national.
H10) Voters who consider voting for more than one party tend to vote more for the most
sympathized with (dislike–like scale) party in the EP elections than in the national elections.
We tested this hypothesis using the same stacked dataset as previously, including all
respondents considering more than one party. Party choice was coded as 1 if the party
received the final vote and 0 otherwise. The party most sympathized with is based on a
18
dislike–like scale ranging from 0 to 10. The results for both elections are displayed in Table
6 and Figure 4 below.
Table 6. Emotional attachment to parties as a secondary decision rule for party choice in
national and EP election campaigns in 2009 and 2010 among the SNES internet campaign
panel participants
Best party
National election
Best party* National
election
Constant
Coefficients
3.31***
0,08
0.30
Robust standard error
0.17
0.054
1.39
−2.82***
0.044
Number of observations
11696
Log pseudolikelihood
−2895.1615
Pseudo R2
0.21
Comment: Combined stacked datatsets from E-panels for 2009 and 2010, meaning that
every respondent occurs as many times as there are parties to choose between. Only
respondents with more than one party in the consideration set are included. The 2009
European election is the reference category, and the 2010 national election is a dummy
variable. The best party is the party with the highest positive value on the dislike–like scale.
Significance levels are marked *= <0.05, ** = <0.01 and ***=<0.001.
19
SNES Internet Campaign Panel 2009 & 2010
0
.2
Pr(Vote)
.4
.6
.8
Predictive Margins of RD with 95% CIs
0
1
Best party
EUP 2009
RD 2010
Figure 4. Emotional attachment to parties as a secondary decision rule for party choice in
national and EP election campaigns in 2009 and 2010 among the SNES internet campaign
panel participants.
The result actually indicates the opposite pattern; that is, a slightly stronger effect of
emotional attachment on party choice in the national election. However, the results are not
significant, which is why the test of the hypothesis is inconclusive.
Conclusions and further research
Elections to the European Parliament are known to be “second order elections.” With the
consideration set approach presented here we are able to compare explanatory factors for
elections on different levels in a way that is more thorough than when the analysis is
restricted to actual party choice. Distinguishing between the consideration stage and the
choice stage opens up further insights into how different explanatory factors work in
different aspects of the decision-making process, not the least the interlaced effects of left–
right vs EU dimensions.
20
The preliminary analyses of the application of the consideration set approach to party
choice in the European election of 2009 and Swedish national election of 2010 have
provided some relevant findings. The analyses indicate a high degree of similarity between
the consideration sets in the two elections respectively, both in terms of how large the
proportion of respondents is that actually considers voting for more than one party, as well
as the size of the consideration sets. Ideological position matters for which parties are
included in the consideration set, with a general pattern that left–right position matters
more than EU issue position at the consideration stage, whilst EU issue position matters
more at the choice stage. However, for the Centre Party and the Liberal Party, EU issue
position actually matters more at the consideration stage than at the choice stage. We have
also confirmed that left–right proximity matters more in relation to which parties are being
considered to vote for in the national election than in the election to the European
Parliament, and that the same goes for the actual vote decision. Due to limited data at this
stage in the project, we have not been able to discern whether this means that proximity
between the voters’ and the parties’ ideological positions on European issues matters more.
Finally, we could not confirm the hypothesis that sincere voting was more decisive at the
choice stage in the EP election than in the national election.
The consideration set model approach, as applied here, confirms the value of distinguishing
between the consideration and the choice stage in the decision-making process about party
choice. It enabled us to isolate the primary from the secondary decision criteria, which is
vital, not least when comparing European with national elections. Only by doing so will we
be able to fully understand how voters decide what party to vote for.
21
Appendix: Table A1. The content of party set for EUP 2009
# of
Parties in
Set
Content of Party Set
Electoral
Support
(percentage)
Number of
Respondents
Ranking of the 10
Most Common
Combinations
0
0-parties (including "do not know")
8.5
140
1
--
--
--
--
--
--
--
--
--
--
M
11
182
#1
1
--
--
--
--
--
--
--
--
--
CD
--
2.1
35
#9
1
--
--
--
--
--
--
--
--
C
--
--
1
17
1
--
--
--
--
--
--
--
Lib
--
--
--
3.7
62
#6
1
--
--
--
--
--
--
SD
--
--
--
--
2.1
34
#10
1
--
--
--
--
--
JL
--
--
--
--
--
0.9
15
1
--
--
--
--
G
--
--
--
--
--
--
4.1
68
#5
1
--
--
--
PP
--
--
--
--
--
--
--
5.4
90
#4
1
--
--
S
--
--
--
--
--
--
--
--
6.3
104
#2
1
--
FI
--
--
--
--
--
--
--
--
--
1
17
1
L
--
--
--
--
--
--
--
--
--
--
2.5
41
40.1
665
1
#8
2
--
--
--
--
--
--
--
Lib
--
--
M
5.5
91
#3
2
--
--
--
PP
--
--
--
--
--
--
M
3.1
52
#7
2
--
--
S
--
G
--
--
--
--
--
--
2.1
34
#10
2
L
--
S
--
--
--
--
--
--
--
--
2
33
2
--
--
--
PP
G
--
--
--
--
--
--
1.5
25
2
L
--
--
--
G
--
--
--
--
--
--
1.5
24
2
--
--
--
--
--
--
--
--
--
CD
M
2
Other 2-party combinations
1.2
19
14.3
238
31.2
516
3
--
--
--
--
--
--
--
Lib
C
--
M
1.8
29
3
--
--
--
PP
--
--
--
Lib
--
--
M
1.3
22
3
L
--
S
--
G
--
--
--
--
--
--
1.2
19
3
L
--
--
PP
G
--
--
--
--
--
--
0.9
15
3
--
--
--
--
--
--
--
Lib
-
CD
M
0.9
15
3
L
FI
--
--
G
--
--
--
--
--
--
0.8
13
3
Other 3-party combinations
7
117
13.9
230
4
--
--
--
--
--
--
--
Lib
C
CD
M
0.5
8
4
L
--
S
PP
G
--
--
--
--
--
--
0.4
6
4
L
FI
S
--
G
--
--
--
--
--
--
0.4
6
4
Other 4-party combinations
3.1
52
4.4
72
1.8
30
100
3136
5-8
5-8-party combinations
22
Appendix: Table A2. The contents of party set for RD 2010
# of
Parties in
Set
Electoral
Support
(percentage)
Content of Party Set
Number of
Respondents
Ranking of the 10
Most Common
Combinations
0
0-parties (including "do not know")
5.2
686
1
--
--
--
--
--
--
--
--
--
--
M
11
1451
1
--
--
--
--
--
--
--
--
--
CD
--
1.3
175
1
--
--
--
--
--
--
--
--
C
--
--
1.1
143
1
--
--
--
--
--
--
--
Lib
--
--
--
3.2
424
#8
1
--
--
--
--
--
--
SD
--
--
--
--
4.8
629
#5
1
--
--
--
--
--
JL
--
--
--
--
--
0
2
1
--
--
--
--
G
--
--
--
--
--
--
4.8
632
1
--
--
--
PP
--
--
--
--
--
--
--
1.1
144
1
--
--
S
--
--
--
--
--
--
--
--
7.9
1034
1
--
FI
--
--
--
--
--
--
--
--
--
0.5
61
1
L
--
--
--
--
--
--
--
--
--
--
1
4.2
555
39.9
5250
#1
#4
#2
#6
2
--
--
--
--
--
--
--
Lib
--
--
M
5.2
676
2
--
--
S
--
G
--
--
--
--
--
--
3.7
489
#7
2
L
--
S
--
--
--
--
--
--
--
--
3
388
#10
2
L
--
--
--
G
--
--
--
--
--
--
2.5
331
2
--
--
--
--
--
--
--
--
--
CD
M
2.2
282
2
--
--
--
--
--
--
SD
--
--
--
M
1.7
222
2
--
--
--
--
--
--
--
--
C
--
M
1.2
154
2
Other 2-party combinations
9.9
1363
29.4
3905
3
L
--
S
--
G
--
--
--
--
--
--
3
391
3
--
--
--
--
--
--
--
Lib
C
--
M
1.8
234
3
--
--
--
--
--
--
--
Lib
--
CD
M
1.7
217
3
L
FI
--
--
G
--
--
--
--
--
--
1.4
180
3
--
FI
S
--
G
--
--
--
--
--
--
0.7
85
3
--
--
--
--
--
--
--
--
C
CD
M
0.6
78
3
Other 3-party combinations
7.6
1000
16.8
2185
4
L
FI
S
--
G
--
--
--
--
--
--
1
125
4
L
--
S
PP
G
--
--
--
--
--
--
0.4
37
4
--
--
--
--
G
--
--
Lib
C
M
0.3
33
4
Other 4-party combinations
2.6
339
4.3
534
5
5-party combinations
1.2
153
6–8-party combinations
0.3
39
100
12752
6-8
#3
#9
23
References
Bengtsson, Å., Hansen, K. M., Narud, H.-M., Hardarson, O., & Oscarsson, H. (2013). The
Nordic Voter: Myths of Exceptionalism. London: ECPR Press.
Carmines, E. G., & Stimson, J. A. (1989). Issue Evolution. Race and the Transformation of
American Politics. Princeton: Princeton University Press.
Dahlberg, S., Lindholm, H., Lundmark, S., Oscarsson, H., & Åsbrink, R. (2010). The 2010
Internet Campaign panel. The Multidisciplinary Research on Opinion and Democracy
(MOD) Group Working paper 2010:3. Department of Political Science,. University of
Gothenburg. Gothenburg.
Dahlberg, S., Lindholm, H., Lundmark, S., Oscarsson, H., & Åsbrink, R. (2011). The 2010
Internet Campaign Panel Multidisciplinary Opinion and Democracy Research Group.
Working Paper 2011:1. University of Gothenburg: MOD.
Dalton, R. (2013). The apartisan American. Dealignment and changing electoral politics.
Thousand Oaks: CQ Press.
Dalton, R. J., & Klingemann, H.-D. (2007). Oxford handbook of political behavior. Oxford ;
New York: Oxford University Press.
Hobolt, S. B., & Spoon, J.-J. (2012). Motivating the European voter: Parties, issues and
campaigns in European Parliament elections. European Journal of Political Research,
51, 701-727.
Hobolt, S. B., & Wittrock, J. (2011). The second-order election model revisited: An
experimental test of vote choices in European Parliament elections. Electoral
Studies, 30, 29-40.
Holmberg, S., & Oscarsson, H. (2004). Väljare. Svenskt väljarbeteende under 50 år. [Voters.
Fifty Years of Voting Behaviour in Sweden]. Stockholm: Norstedts.
Marsh, M. (1998). Testing the Second-Order Election Model after Four European Elections.
British Journal of Political Science, 0(4), 591-607.
Oscarsson, H. (2004). Verkligen väljande väljare [Virtually Voting Voters]. In S. Holmberg &
H. Oscarsson (Eds.), Väljare. Svenskt väljarbeteende under 50 år [Voters. Fifty Years
of Swedish Voting Behavior]. Stockholm: Norstedts Juridik AB.
Oscarsson, H. (2009). Identifying Voters Consideration Sets. . Paper presented at the ECPRconference, Potsdam, Germany, .
Oscarsson, H., Gilljam, M., & Granberg, D. (1997). The Concept of Party Set. A Viable
Approach or Just Another Way to Slice the Same Cheese? Göteborg University:
Department of Political Science.
Oscarsson, H., & Holmberg, S. (2008). Regeringsskifte. Väljarna och valet 2006. [Government
Turnover. Voting behavior in the 2006 Election]. Stockholm: Norstedts Juridik.
Paap, R., Nierop, E. v., Heerde, H. J. v., Wedeld, M., Fransesa, P. H., & Alseme, K. J. (2005).
Consideration sets, intentions and the inclusion of ‘‘don’t know’’ in a two-stage
model for voter choice. International Journal of Forecasting, 21, 53-71.
Popkin, S. L. (1991). The Reasoning Voter: Communication and Persuasion in Presidential
Campaigns. Chicago: University of Chicago Press.
Reif, K., & Schmitt, H. (1980). Nine Second-Order Elections - A Conceptual Framework for
the Analysis of European Election Results. European Journal of Political Research, 8,
3-44.
24
Roberts, J. H., & Lattin, J. M. (1991). Development and Testing of a Model of Consideration
Set Composition. Journal of Marketing Research, 28(4), 429-440.
Roberts, J. H., & Lattin, J. M. (1997). Consideration: Review of Research and Prospects for
Future Insights. Journal of Marketing Research, 34(3), 406-410.
Schmitt, H. (2005). The European Parliament Elections of June 2004: Still Second-Order?
West European Politics, 28(3), 650-679.
Shocker, A. D., Ben-Akiva, M., Bocara, B., & Nedungadi, P. (1991). Consideration Set
Influences on Consumer Decision-Making and Choice: Issues, Models, and
suggestions. Marketing Letters, 2(3), 181-197.
Sniderman, P. M., Brody, R. A., & Tetlock, P. E. (Eds.). (1991). Reasoning and Choice.
Explorations in Social Psychology. Cambridge: Cambridge University Press.
Steenbergen, M. R., & Hangartner, D. (2008). Political Choice Sets in Multi-Party Elections.
Paper presented at the The politics of change, Amsterdam.
Steenbergen, M. R., Hangartner, D., & Vries, C. E. d. (2011). Choice under Complexity: A
Heuristic-Systematic Model of Electoral Behavior. Paper presented at the Annual
Meeting of the Midwest Political Science Association, Chicago.
van der Brug, W., & Eijk, C. v. d. (Eds.). (2007). European Elections & Domestic Politics.
Lessons from the past and scenarios for the future. Notre Dame: Notre Dame
University Press.
van der Eijk, C., & Franklin, M. N. (2007). The sleeping giant: Potential for political
mobilization of disaffection with European integration. In W. van der Brug & C. Van
der Eijk (Eds.), European Elections & Doemstic Politics. Lessons from the past and
scenarios for the future. (pp. 189-208). Notre Dame, Indiana: University of Notre
Dame Press.
Wilson, C. J. (2008). Consideration Sets and Political Choices: A Heterogeneous Model of
Vote Choice and Sub-national Party Strength. Political Behavior, 30, 161-183.
Vries, C. d., Steenbergen, M. R., & Hangartner, D. (2009). A Choice Set Modeling Approach to
EU Issue Voting. . Paper presented at the EUSA Conference panel "Transnational
Politics and the Vote", Los Angeles USA.
Vries, C. E. d. (2007). Sleeping Giant: Fact or Fairytale? Eurpean Union Politics, 8(3), 363-385.
Vries, C. E. d., Brug, W. v. d., Egmond, M. H. v., & Eijk, C. v. d. (2011). Individual and
contextual variation in EU issue voting: The role of political information. Electoral
Studies, 30, 16-28.
Zaller, J. (1992). The Nature and Origins of Mass Opinions. Cambridge: Cambridge University
Press.
25
26